Deep metagenomics uncovers functional adaptations and pathogenic risks in the gut microbiome of Antarctic fur seals (Arctocephalus gazella)

Abstract

The Antarctic fur seal (Arctocephalus gazella) plays a key role in the Antarctic marine ecosystem by regulating krill, fish, and cephalopod populations through selective foraging, promoting Southern Ocean productivity via excretion, and influencing coastal island ecosystems during breeding season. Despite the importance of the gut microbiota in reflecting diet, health, and environmental adaptation, the gut microbiome of the Antarctic fur seal remains poorly characterized. To address this gap and evaluate its potential as a bioindicator of Antarctic marine environmental health, we employed shotgun metagenomics and 16S rRNA amplicon sequencing on fresh fecal samples collected from four Antarctic fur seals (designated S59, S62, S63, and S64) at King George Island, Western Antarctica. Despite inter-individual variation, both approaches identified Bacillota as the dominant phylum but showed genus-level discrepancies, with Fusobacterium prevailing in metagenomes and Clostridium in 16S amplicons. Viral communities constituted up to 5.3% of the microbiome, including an immunodeficiency-associated Lentivirus. Chitin-degrading capacity was ubiquitous, consistent with the host’s krill-based diet. Metagenome-assembled genomes (MAGs) resolved distinct taxonomic contributions to discrete steps of chitin hydrolysis, suggesting that complete depolymerization requires metabolic cross-feeding among functionally complementary taxa. Notably, Helicobacter MAGs were abundant in individual S62, suggesting potential pathogenicity. Additionally, 16 antibiotic resistance gene types were detected, with bacitracin, polymyxin, and multidrug resistance dominating the resistome. These findings not only elucidate the community composition, functional potential, and ecological adaptation of the Antarctic fur seal gut microbiota but also establish a comprehensive baseline for assessing environmental change and human impacts on the Antarctic marine ecosystem, thereby offering valuable scientific data and methodological insights for the conservation of polar marine mammals.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40793-026-00919-2.

Background

The Antarctic fur seal (Arctocephalus gazella) is a keystone species in the Antarctic and subantarctic ecosystems, serving as an indicator of ecosystem health [1]. This species has a circumpolar distribution, with most breeding colonies found in islands close to the Antarctic Polar Front [2]. As a top predator primarily feeding on Antarctic krill (Euphausia superba) which makes up about 95% of its diet, the Antarctic fur seal forms a critical trophic link between primary producers and apex predators [3–5]. Its feces enriched in nitrogen can alter soil nutrient composition, promoting the growth of specific microbial communities, and significant impacting nutrient cycling in Antarctic terrestrial ecosystems [6, 7]. In the Southern Ocean, where iron limitation constrains phytoplankton growth, seals serve as biogeochemical vectors by consuming iron-enriched prey and subsequently excreting bioavailable iron, generating localized fertilization zones that enhance primary production [8]. Concurrently, highest Hg concentrations were detected in fur seal excreta compared to penguins and other seals, leading to increase of Hg concentrations in habitat soils and marine sediments of distribution range [9]. This distribution and dietary specialization underscore the Antarctic fur seal’s role as not only a predator but an indicator species for the health of the Antarctic marine ecosystem by reflecting the impacts of climate change, environmental pollution, and human activities. The population of Antarctic fur seals has undergone extreme decline to almost extinction due to sealing industry in last century, and later recovered due to protection. However, it is currently experiencing a decline due to climate change [10]. Their population dynamics and physiological conditions serve as sensitive bioindicators for monitoring ecosystem health and anthropogenic impacts in polar regions, including climate change, hunting, and krill fishery pressure [11, 12]. Consequently, understanding the ecological physiology of A. gazella is indispensable for Antarctic biodiversity conservation and environmental monitoring initiatives.

The gut microbiome has emerged as a focal point in ecological and physiological research, providing critical insights into host health, metabolic processes, and evolutionary adaptations [13, 14]. Antarctic pinnipeds exhibit pronounced dietary segregation: Lobodon carcinophaga feeds almost exclusively on krill; Hydrurga leptonyx, Mirounga leonina, and Leptonychotes weddellii are primarily piscivorous, large female leopard seals preferentially prey on fur seal pups and penguins; and Ommatophoca rossii feeds predominantly on cephalopods; A. gazella displays marked geographic variation, with South Georgia populations dominated by krill consumption [15], whereas Macquarie Island populations rely mainly on fish [16]. Such interspecific and spatial differences in prey composition are likely to drive corresponding differences in gut microbiome composition [17–19]. At present, knowledge of the gut microbiota of the krill-feeding A. gazella remains limited and mainly relies on 16S rRNA gene amplicon sequencing. For example, using 16S rRNA amplicon sequencing, Botsidou [19] reported that the gut microbiota of Antarctic fur seals is dominated by Bacillota, with notable contributions from Pseudomonadota, Bacteroidota, and Fusobacteriota, whereas metagenomic analyses suggested Fusobacteriota as the dominant phylum [20]. Studies of other Antarctic pinnipeds further reveal broadly similar gut microbiome compositions at the phylum level, as exemplified by wild southern elephant seals and leopard seals, both of which are characterized by the dominance of Bacillota [17]. However, gut microbial richness was significantly higher in southern elephant seals than in wild leopard seals, with only six operational taxonomic units (OTUs) shared between captive and wild leopard seals [17]. Similarly, Kim [21] demonstrated that despite similar phylum-level patterns in Weddell seals and southern elephant seals, their amplicon sequence variants (ASVs) showed virtually no overlap. Collectively, these studies suggest that the gut microbial community is strongly shaped by host species, diet, and environmental conditions.

Beyond bacterial community structure, recent studies have revealed additional microbial and anthropogenic signals in Antarctic pinniped guts. Novel viral species belonging to Circovirus (werosea circovirus), Cyclovirus (werosea cyclovirus), and Anelloviridae (torque teno Leptonychotes weddellii virus-1 and -2) have been detected in L. weddellii [22, 23]. Notably, the concentrations of Escherichia coli and Clostridium perfringens detected in Weddell seal feces were comparable to those reported in untreated sewage discharged from McMurdo Station [24], and human-associated E. coli clonal groups have been identified in Antarctic pinnipeds [25]. In addition, approximately 80% of bacterial isolates from Antarctic marine mammal samples exhibited antibiotic resistance [26]. Recent investigations of Antarctic fur seal gut viromes have further demonstrated associations with dietary habits and interactions with human activities [20, 27, 28]. For marine mammals, particularly in polar regions experiencing compounded pressures from climate change and anthropogenic disturbance, the gut microbiome represents a dynamic interface linking host biology with ecosystem processes. Therefore, to better elucidate the relationships between the gut microbiota of A. gazella and diet, health, and environmental factors, there is an urgent need to expand metagenomic investigations of the A. gazella gut microbiome.

This study investigated the gut microbiota of Antarctic fur seals (A. gazella) using shotgun metagenomics and 16S rRNA amplicon sequencing, gaining novel insights into dietary adaptations, health status, and anthropogenic impacts. As climate change and human activities increasingly disrupt polar ecosystems, these findings offer an evidence-based foundation for developing targeted conservation strategies to mitigate environmental pressures on seal populations.

Methods

Sample collection and preliminary processing

During the 36th Chinese National Antarctic Research Expedition, fresh fecal samples of A. gazella were collected into sterile bags immediately following defecation by four distinct individuals, thereby ensuring that the samples were obtained promptly after excretion. The surface of the fecal matter in contact with the ground was removed to avoid soil contamination. The sampling sites were within a 1-km range on the Fildes Peninsula, King George Island, West Antarctica (Fig. 1a). The samples were stored at -80 ℃ after collection for subsequent analyses.

Fig. 1.

Sampling location, host species, and fecal sample characterization. a The sampling sites were within a 1-km range on the Fildes Peninsula, King George Island, Antarctica, as indicated by a red dot on the map. b Images of A. gazella observed in the field and their corresponding fecal samples. c Residues extracted from the fecal samples, primarily composed of Antarctic krill exoskeletons and suspected fish bones

Thawed fecal samples were sequentially rinsed with Milli-Q water to isolate indigestible prey components (e.g., Antarctic krill shells, otoliths and cephalopod beaks) and analyzed visually and microscopically. The resulting hard parts were fixed in 30% neutral-buffered formaldehyde solution at 4 ℃ for 24 h to prevent microbial degradation prior to morphological analysis.

DNA extraction and sequencing

DNA from fecal samples was extracted using the DNeasy PowerSoil Kit (Qiagen) according to the manufacturerer’s instructions. The integrity and RNA contamination of the gDNA were checked through 1% agarose electrophoresis. The quality and quantity of the gDNA were evaluated using a NanoDrop™ spectrophotometer and a Qubit Fluorometer (Invitrogen), respectively.

For 16S rRNA amplicon sequencing, the hypervariable V3-V4 region of the 16S rRNA gene was amplified using primers 338F (5’-ACTCCTACGGGAGGCAGCA-3’) and 518R (5’-GGACTACHVGGGTWTCTAAT-3’). The PCR products were recovered using Vazyme VAHTS™ DNA Clean Beads (Vazyme Biotech) and quantified with the Quant-iT PicoGreen dsDNA Assay Kit (Thermo Fisher Scientific) on a microplate reader (BioTek, FLx800). The library was subsequently constructed using the TruSeq Nano DNA LT Library Prep Kit (Illumina) and subjected to paired-end sequencing on the NovaSeq6000 platform (Illumina) at Shanghai Personal Biotechnology Co., Ltd. (Shanghai, China).

For metagenome sequencing, gDNA was fragmented by sonication using the Covaris instrument and a 300–400 bp DNA library was constructed using BGI Optimal DNA Library Prep Kit. Sequencing was performed on the DNBSEQ-T7 platform with PE150 strategy at BGI (BGI-Wuhan, China), generating 87–100 million raw reads (13.4–15.0 Gb) per library.

Community analysis via 16S rRNA gene sequencing

The 16S rRNA gene amplicon data were processed using the QIIME2 platform [29] following established microbial ecology workflows. Raw sequences were first subjected to adapter and primer trimming, followed by demultiplexing based on sample-specific barcodes. Initial quality control and sequence refinement were then performed using the DADA2 plugin [30] with the default independent pooling strategy, which facilitated chimera removal and error correction to generate ASVs. For phylogenetic classification, we employed a machine learning approach using the SILVA v138_99 [31] reference database with a pre-optimized Naïve Bayes classifier (q2-feature-classifier) [32] applying a confidence threshold of ≥ 0.7 for reliable taxonomic assignments.

Shotgun metagenomic assembly and functional potential

Raw sequencing data were subjected to quality control: potential adapter sequences were removed by Cutadapt (v1.17) [33], and low-quality reads were further filtered with fastp (v0.20.0) [34] applying the default Phred quality score threshold (Q15) to ensure only high-quality sequences were retained for subsequent analyses. Following quality control, each library generated between 83 and 94 million high-quality reads (12.5–14.2 Gb). Filtered sequences were assembled using MEGAHIT (v1.2.9) [35] with a single-sample paired-end assembly strategy (parameters: –k-min 21, –k-max 141, –k-step 10). After assembly, open reading frames (ORFs) were predicted from assembled contigs using Prodigal (v2.6.3) [36]. The predicted gene sequences were then combined and dereplicated to generate a high-quality gene set. The ORFs from each sample and the mixed assembly were clustered using CD-HIT (v4.8.1) [37] (parameters: -c 0.95, -M 0, -T 80, -n 10, -d 0, -aS 0.9, -g 1, -sc 1, -sf 1), and the longest sequence from each cluster was selected to form a non-redundant gene catalog (unigenes). Clean reads from each sample were then mapped to this catalog using BWA-MEM (v0.7.17-r1188) [38] to calculate the abundance of each unigene.

To maximize taxonomic annotation rate and improve biological interpretability, the unigenes were compared against the NCBI-NR database using DIAMOND (v2.0.15) [39], and gene abundance data were integrated across taxonomic ranks. Functional annotation was performed by searching unigenes against the Kyoto Encyclopedia of Genes and Genomes (KEGG) [40], Comprehensive Antibiotic Resistance Database (CARD) (identity ≥ 80%, query coverage ≥ 25%, e-value ≤ 1e-7) [41], Virulence Factors DataBase (VFDB) [42], and Carbohydrate-Active enZYmes Database (CAZy) [43], and functional gene annotations were integrated with the abundance data to provide a detailed profile of functional gene composition and their relative abundances. The abundance of ARGs was normalized to the 16S rRNA gene copy number at the ARG type level using ARGs-OAP [44]. Metagenome-assembled genomes (MAGs) were recovered with a hybrid binning strategy implemented using MetaWRAP (v1.3.2) [45], integrating consensus results from metaBAT2, MaxBin2, CONCOCT, and metaBAT1 algorithms. MAGs were refined based on completeness > 80% and contamination < 10%, as assessed by CheckM [46]. Refined MAGs were further dereplicated using dRep (v3.2.2) [47] with default parameters, and taxonomically classified using GTDB-Tk (v2.1.0) with the Genome Taxonomy Database (GTDB r207) [48]. Phylogenetic relationships of dereplicated MAGs were reconstructed via IQ-TREE2 (v2.1.3) [49] under the optimal evolutionary model (LG + F + R10) identified by ModelFinder [50]. ORFs were predicted for each MAG using Prodigal (v2.6.3) in metagenomic mode, and gene abundance was calculated with BWA-MEM (v0.7.17) as above.

Data analysis

Data analysis and visualization were performed using R software (V3.6.0) [51]. Statistical analyses were carried out using standard methods from the base package. For the 16S rRNA gene sequencing data, taxonomic and ASV abundance tables were derived directly from read counts. Relative abundances were calculated by dividing the read counts of each taxon or ASV by the total read counts per sample. For shotgun metagenomics data, gene abundances were normalized using Transcripts Per Million (TPM) via the rarefy_even_depth function in the phyloseq package (v1.36.0) [52] after aligning clean reads to the non-redundant gene catalog; taxonomic and functional relative abundances were then obtained by summing the TPM values of genes assigned to each taxon or functional category. α-diversity indices (Shannon, Simpson, Species Richness, Pielou) were subsequently calculated using the diversity and specnumber functions from the vegan package. β-diversity metrics (Bray–Curtis) [53] were also calculated from genus abundance tables using the vegan package. Principal coordinates analysis (PCoA) was performed using the Bray–Curtis distance and the classical multidimensional scaling (cmdscale) algorithm via the vegan package. Similarly, non-metric multidimensional scaling (NMDS) was also conducted using the Bray–Curtis distance and the vegan package, employing an iterative sorting algorithm to minimize the stress value, calculating the distances between samples and exploring changes in community structure. Multivariate statistical comparisons of community structure were performed using PERMANOVA [54] with 999 permutations in the vegan package. Graphical representations were generated using the ggplot2 package, with color schemes provided by the ggsci package to enhance the presentation of the results.

Results

Dietary correlation with fecal color and residual material

Fresh fecal samples collected from four individual A. gazella (S59, S62, S63 and S64) at the coast of Fields Peninsula exhibited bright red coloration, indicating Antarctic krill consumption (Fig. 1). Visual inspection revealed heterogeneous residue morphologies: translucent fragments, orange particles, and reddish residues resembling the original fecal coloration (Fig. 1c). Microscopic analysis confirmed that the predominant residues consisted of exoskeletal fragments of Antarctic krill (E. superba), including diagnostic structures such as tail segments, thoracopods, and pleopods. In addition to krill remains, a minor fraction of the residues resembled fish bones. This observation reflected the dietary habits of A. gazella, with Antarctic krill as the primary food source and fish serving as a secondary contributor [55].

Microbial structure based on 16S rRNA amplicon sequencing

After quality control, over 100,000 clean reads were retained per sample, providing a robust dataset for microbiome analyses. A total of 11 phyla were detected, with Bacillota (73.3 ~ 91.2%) as the predominant phylum across all four samples. Other major phyla, including Fusobacteriota (9.8% ~ 24.3%), Bacteroidota (1.1% ~ 3.8%), and Pseudomonadota (0.2% ~ 2.8%), were also consistently represented across all samples, albeit with varying relative abundances (Fig. 2a). Specifically, Deferribacterota was uniquely identified in sample S63, Gemmatimonadota was found solely in samples S62 and S63, and Saccharibacteria was exclusively present in samples S62 and S64.

Fig. 2.

Microbial community structure and diversity in four samples based on 16S rRNA amplicon sequencing. a Relative abundance of microbial phyla. b Relative abundance of the top 20 microbial genera. c Venn diagram showing the number of shared and unique genera across the four samples

At the genus level (Fig. 2b-c), 148 genera were detected, with 30 genera shared across all four samples. The shared genera included 19 known genera and 11 unclassified genera from families such as Fusobacteriaceae, Lachnospiraceae, Ruminococcaceae, and Erysipelotrichaceae (Table S1). Among these, Clostridium emerged as the most abundant genus, accounting for 43.7 ~ 72.5% The second-most abundant genus varied by sample: unclassified Fusobacteriaceae (6.7–23.6%) dominated in most samples, except in S63, where unclassified Lachnospiraceae (14.6%) outranked it (13.1%). In addition to Clostridium, other typical gut genera were frequently found, including Dorea, Lactobacillus, Bacteroides, Enterococcus, Ruminococcus, Streptococcus, Blautia, Faecalibacterium, and Butyricicoccus were also present in all samples, albeit at much lower relative abundances, ranging from ~ 0.01% to 4.96%. Genera well-adapted to polar environments, such as Psychrobacter [56], Polaromonas [57], and Deinococcus [58] were also detected in 3 ~ 4 samples.

According to the Shannon, Simpson, Species Richness, and Pielou indices, sample S63 exhibited the highest diversity while S64 had the lowest (Table S2). Both PCoA and NMDS analyses demonstrated close clustering of samples S62 and S59, reflecting similar microbial community structures, whereas S63 and S64 were distantly separated in the ordination space, suggesting substantial compositional differences (Bray–Curtis distance = 0.83 ± 0.23; Fig. 2d-e; Fig. S1a-b).

Metagenomics-based community characterization

1. Overall characterization

Metagenomic sequencing of the A. gazella gut microbiome revealed a pronounced dominance of bacteria across all samples, constituting over 94.5% of the microbial community (Fig. 3a). Viruses represented the second-most abundant domain, with relative abundances ranging from 2.1% (S62) to 5.3% (S64). The proportion of eukaryote and archaea was lower, averaging 0.2% and 0.05%, respectively. A total of 2,534 genera were identified, comprising 437 unclassified genera and 2,097 known genera. Notably, 1,522 genera from all the domains were shared across all four samples. Each sample also harbored various number of unique genera, for instance, sample S59 exhibited the highest number (119) of unique genera, whereas sample S64 had only 65. Both PCoA and NMDS ordination demonstrated clear segregation of the four samples into discrete clusters, with approximately equivalent pairwise dissimilarities between all sample groups (Bray–Curtis distance = 0.84 ± 0.20; Fig. S1c-d).

Fig. 3.

Microbial community structure and diversity revealed by metagenomic sequencing. a Relative abundance of microbial communities across four domains: Bacteria, Viruses, Eukaryota, and Archaea. b Relative abundance of the top 20 microbial phyla. c Relative abundance of the top 20 microbial genera. d Venn diagram showing the number of shared and unique genera across the four samples

• (2)Bacterial composition

Consistent with 16S rRNA amplicon data, Bacillota, Fusobacteriota, Bacteroidota and Pseudomonadota were the top bacterial phyla in the metagenomes (Fig. S3a, Table S3). Bacillota predominated in samples S59 (43.0%) and S62 (57.7%), while Bacteroidota (37.0%) and Fusobacteriota (39.5%) were most abundant in S63 and S64, respectively. In contrast to 16S rRNA amplicon data, the metagenomes revealed numerous minor bacterial phyla, including many candidate phyla (each < 0.1% and collectively < 1% of the total), thereby increasing the total bacterial phyla to 106 (including unclassified and candidatus phyla).

The identified bacterial phyla encompassed 2,269 genera, among which 335 were unclassified, 247 were candidatus, and 1,687 were known genera (Table S4). Fusobacterium (21.5 ~ 37.9%) dominated across all samples. Bacteroides was the second abundant genus in samples S59, S63 and S64, ranging from 5.6% to 18.1%, while unclassified Oscillospiraceae genus ranked second in S62 (10.2%). Butyricicoccus was the third abundant genus in S62, accounting for 6.4%, much higher than in the other three samples (0.5 ~ 1.4%). Most of the remaining genera were rare, with less than 1% in abundance.

A total of 107 bacterial genera were detected by both 16S rRNA amplicon sequencing and metagenomics, representing 76.4% (113/148) of genera identified via 16S sequencing and 5.0% (113/2,269) of those from metagenomes. The shared genera exhibited markedly different abundance patterns between the two methods (Table S5, Fig. 3), as detailed above and further analyzed in the Discussion.

Meanwhile, several pathogenic or opportunistic pathogenic genera were detected, including Streptococcus, Helicobacter, Enterococcus, Staphylococcus, Acinetobacter, Salmonella, Clostridium, Campylobacter, Legionella, Yersinia, Klebsiella, and Listeria etc. (Tables S4, S5).

• (3)Viral composition

The high proportion of virus (2.1% to 5.3%), particularly the pronounced peak in S64, may indicate transient phage activity or latent infection states. We found 8,460 viral contigs after de novo assembly. Single-strand DNA viruses (ssDNA) were the most diverse among viral contigs (56.6%), followed by double-strand DNA phage (dsDNAphage) (43.4%). Four known viral phyla (Uroviricota, Phixviricota, Hofneiviricota and Pisuviricota) and an unclassified phylum were detected in the metagenomes. Uroviricota was the predominant viral phyla, accounting for 90.2 ~ 95.1% in the viral communities, corresponding to 2.0 ~ 4.8% in the total metagenomes. More than eleven viral families (Anelloviridae, Bacciladnaviridae, Caliciviridae, Circoviridae, Guelinviridae, Kyanoviridae, Microviridae, Nodaviridae, Pachyviridae, Parvoviridae, and Picornaviridae) were detected. Of the 67 viral genera detected (54 classified, 13 unclassified), over 90% were unclassified Caudoviricetes bacteriophages. Lentivirus—an immunodeficiency-associated genus, was detected exclusively in sample S59, and represented the 12th most abundant viral genus (Table S4, Fig. S3b).

• (4)Eukaryotic composition

Although eukaryotes represented only around 0.2% in the metagenomes, they comprised a diverse assemblage, including protozoa (e.g., Apicomplexa, Ciliophora, Fornicata), metazoan (e.g., Nematoda, Arthropoda, Chordata, Mollusca), plants (e.g., Streptophyta), fungi (Basidiomycota, Mucoromycota, Ascomycota), and algae (e.g., Chlorophyta, Bacillariophyta). Up to 149 genera were identified, the plant genus Lupinus was the most abundant across all samples (Fig. S3c). Moreover, trace amounts of other plant-derived genera were detected, including both wild species (Artemisia and Tanacetum) and cultivated crops (Cajanus, Citrus, Pisum, Brassica). Several metazoan parasites were identified, particularly nematode genera Anisakis, Trichuris, and Toxocara, which are known to infect marine and terrestrial vertebrates. Of particular ecological relevance was the identification of Prasiola, a psychrophilic filamentous chlorophyte that forms symbiotic associations with fungi in Antarctic lichen communities, which can be frequently found in Fields Peninsula.

• (5)Archaeal composition

The abundance of Archaea was even lower than Eukaryote. Only 15 archaeal phyla were detected, including Euryarchaeota, Thermoproteota, Nitrososphaerota, Nanoarchaeota, ten other Candidatus phyla, and an unclassified phylum. Most of the known archaeal genera were methanogens, including Methanosphaera, Methanobrevibacter, Methanosarcina, Methanobacterium, Methanococcus, Methanocorpusculum and Methanogenium (Table S4). The first three were commonly found in gut microbiomes [59–62]. Other groups most likely found in marine environments were present, such as Candidatus Poseidoniales and Candidatus Bathyarchaeota. Some groups unlikely found in gut microbiomes were also detected such as Thermofilum, Thermoprotei and Thermoplasmata.

Functional overview

Functional annotation utilizing the KEGG pathway database revealed a total of eight and 55 metabolic pathways at level 1 and level 2, respectively. The identified pathways span several key biological processes, such as genetic information processing, signaling and cellular processes, carbohydrate metabolism, and amino acid metabolism (Fig. 4d). Additionally, carbohydrate metabolism (15 pathways, with amino sugar and nucleotide sugar metabolism (0.98%) being the most abundant), amino acid metabolism (14 pathways, dominated by cysteine and methionine metabolism (0.47%)), and lipid metabolism (15 pathways, led by fatty acid biosynthesis (0.38%)) exhibited enrichment of multiple specific pathways. Among these, the pathways most closely associated with the krill-based diet were amino sugar and nucleotide sugar metabolism (0.98%) [63], pyruvate metabolism (0.76%) [64–66], valine, leucine and isoleucine degradation (0.049%) [67, 68], glycerophospholipid metabolism (0.33%) [69, 70], and fatty acid degradation (0.12%) [71]. Sensory system pathway exhibited minimal representation. While minor inter-individual variations were observed, the overall functional profiles showed remarkable conservation in core pathways.

Fig. 4.

Functional characteristics of the microbial community revealed by KEGG-based metagenomic analysis. a Venn diagram showing the number of shared and unique KEGG Orthology (KO) across the four samples. b Heatmap showing log10-transformed relative abundance of level 2 pathways

The Venn diagram analysis based on KEGG Orthology (KO) identified 5,341 shared KOs across all four individuals (Fig. 4a). The intersection between samples S62 and S63 encompassed 5,997 shared KOs, whereas S59 and S64 exhibited the least overlap. Sample-specific uniqueness was also evident: S59 harbored 200 unique KOs, compared to only 30 in S64. These unique KOs, potentially encoding specialized enzymes for niche-specific substrate utilization. Multivariate analyses by both PCoA and NMDS consistently showed homogeneous distribution of samples (Fig. S2a-b).

Genetic analysis related to chitin degradation

Given Antarctic krill’s role as the primary prey of Antarctic fur seals, we specifically investigated microbial chitin degradation pathways, as chitin constitutes the predominant residual material from krill digestion. Metagenomic analysis revealed the coexistence of both oxidative (orange arrows) and hydrolytic (blue arrows) chitin degradation pathways (Fig. 5a).

Fig. 5.

Metagenomic insights into microbial chitin degradation pathways and heatmap of related enzymes and genes in fecal samples. a Chitin degradation pathways, including oxidation (yellow arrows) and hydrolysis (blue arrows). Colored boxes indicate presence of matching sequences in the metagenomic data, and gray boxes indicate absence; gray background with question marks indicate uncertain presence; and asterisks (*) indicate suspected presence based on partial or indirect evidence. b Heatmap showing log10-transformed relative abundance of chitinase-encoding genes across samples

The oxidative pathway, initiated through lytic polysaccharide monooxygenase (LPMO)-mediated chitin oxidation is rarely mentioned in previous studies. The oxidized chitinoligosaccharides, generated through the synergistic action of LPMO and chitinase, were subsequently transported into the periplasmic space via TonB-dependent receptors. Further degradation was performed by OngA, a putative β-hexosaminidase belonging to the GH20 family that has been recently discovered in marine bacteria [72]. The products from OngA were further transported into the cytosol through SSS family sympoters and sequentially degraded by a series of enzymes as listed in Fig. 5. All enzymes required for chitin oxidative degradation were present across all four samples, except LPMO and OngC (2-(amino)-2-deoxy-D-gluconic acid deaminase). In the chitin oxidative degradation pathway, 2-Keto-3-deoxy-D-gluconate-6-phosphate (KDG-6P), the final product, enters the glycolysis pathway for further metabolism.

The hydrolytic pathway is much more complex. Key enzymes such as chitinase and beta-N-acetylhexosaminidase (nagZ), as well as some membrane transporters were consistently found across all samples (Fig. 5a, b). Despite the absence of some enzymes such as N-acetylglucosamine kinase (nagK), the presence of key enzymes suggested the existence of chitin hydrolysis pathway. And the core enzymes exhibited conserved abundance across samples, further affirming basic chitin degradation capacity (Fig. 5b). Certain enzymes varied in abundance, for example, chbP and csxA showed higher abundance in samples S59 and S63 compared to S62 and S64.

The relative abundance and composition of virulence factor genes

A total of 4,897 virulence factor genes (VFGs) were identified by comparing against VFBD database. Various VFGs, directly or indirectly associated with host adaptation and pathogenicity, were detected across the samples, being mainly classified under immune modulation (27.6%), nutritional/metabolic factor (22.1%), adherence (13.5%), motility (8.5%), and effector delivery system (7%) (Fig. S4). The abundance of VFGs in different samples ranged from 6.3 × 10^–2 to 1.3 × 10^–1 copies of VFGs per copy of the 16S rRNA gene, with the highest relative abundance at sample S62. The virulence factors involved in surface adhesion and immune evasion (e.g., Type IV pili, flagellar, capsular polysaccharide, elongation factor EF-Tu, porin), host barrier penetration (e.g., type IV secretion system (T4SS), streptococcal plasmin receptor/glyceraldehyde-3-phosphate dehydrogenase), and nutrient acquisition and stress response (e.g., ABC transporter family, heme biosynthesis and uptake, virulence regulators Response Regulator X3 (Regx3), pyrimidine biosynthesis).

Capsule (> 9%), capsular polysaccharide (> 1%), lipooligosaccharide (LOS, > 4.8%) and lipopolysaccharide (LPS, > 4.5%) were highly present across samples compared to other virulence factors (Fig. 6). T4SS was relatively enriched in S62 while ABC transporters showed the highest abundance in S64. Despite slight differences in virulence factor distribution among the samples, the high degree of similarity suggested that core virulence factors were conserved across Antarctic fur seal gut microbiome.

Fig. 6.

Heatmap showing the log10-transformed relative abundance of the top 20 virulence factors based on the VFDB across samples

The relative abundance and composition of antibiotic resistance genes

A total of 301 antibiotic-resistance genes (ARGs) were identified, with total ARG abundance ranging from 4.7 × 10^–2 to 1 × 10^–1 copies of ARGs per copy of the 16S rRNA gene, with the highest abundance in sample S62. A total of 16 ARG types and 74 ARG subtypes were detected. Of these, nine types were shared by all samples, including antibacterial fatty acid, bacitracin, macrolide-lincosamide-streptogramin, multidrug, mupirocin, pleuromutilin and tiamulin, polymyxin, tetracycline, and vancomycin (Fig. 7). Among them, bacitracin (45.9%) was the most abundant, followed by the resistance against polymyxin (28.3%), multidrug (16.2%), and macrolide-lincosamide-streptogramin (4.7%). Notably, genes conferring resistance to antibiotics relevant to both clinical and agricultural use (e.g., tetracycline, vancomycin, beta-lactam, polymyxin) were widely distributed. Significant inter-individual variations were observed, for instance, fosfomycin was detected only in S63, while aminoglycoside, rifamycin and novobiocin occurred in half of the samples.

Fig. 7.

Heatmap showing the log10-transformed relative abundance of antimicrobial resistance genes based on the CARD database across samples

Analysis of MAGs

A total of 23 dereplicated Metagenome-Assembled Genomes (MAGs) with high quality (contamination < 10%, completeness > 80%) were assembled from the metagenomes, falling into 13 genera of bacteria (Fig. 8). All were detected across four samples, and most of them exhibited the lowest abundance in sample S64. Clostridium exhibited the highest abundance (~ 3.6%), ranking within the top 20 alongside Butyricicoccus (~ 2.3%). Vescimonas (< 0.0087%) was detected at lowest abundance, Angelakisella, Faecalimonas, and Alloprevotella displayed relatively low abundances, ranging between 0.018% and 0.03%. Significant variation in the relative abundance of different MAGs was observed among the samples (Fig. 8b). For example, Butyricicoccus is the most abundant MAG in S62, whereas Faecalimonas accounted for the largest proportion in S64. Some MAGs were unique to specific samples, which may reflect the specific environmental conditions or host-specific factors influencing the microbial community composition.

Fig. 8.

Relative abundance profiles, phylogenetic relationships, and chitin-degrading enzyme distribution of metagenome-assembled genomes (MAGs) across samples. a Relative abundance of 23 high-quality MAGs recovered from the four samples. b Phylogenetic tree of the Helicobacteraceae family based on three MAGs (S62-bin.1, S63-bin.4, and S62-bin.7) and reference genomes. Branch lengths indicate substitutions per site (scale bar: 0.2). c Phylogenetic tree constructed from evolutionary distances of 23 high-quality MAGs (the scale bar of 0.1 representing evolutionary divergence), along with the heatmap showing the distribution of chitin degradation enzymes across MAGs based on mapped read counts

Taxonomic annotation of the three MAGs (S62-bin.7, S63-bin.4, and S62-bin.1) confirmed their classification within the family Helicobacteraceae (Fig. 8a). Phylogenetic analysis revealed that S62-bin.7 and S63-bin.4 formed a tight cluster on a shared branch, exhibiting closest affinity to Wolinella succinogenes, which was originally isolated from the bovine rumen [73]. S62-bin.1 clustered near Helicobacter trogontum and Helicobacter bilis, which were initially isolated from the intestines of rats and mice, respectively [74, 75]. Notably, all three MAGs were taxonomically assigned to the genus Helicobacter and exhibited a close phylogenetic relationship to the well-characterized enteric pathogen Helicobacter pylori, although they did not form a monophyletic cluster (Fig. 8b).

Among the analyzed MAGs (Fig. 8c), Clostridium encoded the highest number of chitin-degrading enzymes (17 enzymes), closely followed by Holdemania (16 enzymes). However, both lacked chitinase and lpmo, which are key enzymes initiating chitin hydrolysis and oxidation. In contrast, Faecousia, Massiliimalla and Angelakiella had chitinase that suggest the potential oxidative pathway for chitin degradation in them. Certain MAGs, such as those from Helicobacter, contained the fewest chitin-degrading enzymes, and the enzymes identified in these MAGs were predominantly associated with membrane transport functions. Notably, the number of chitin-degrading enzymes varied among MAGs within the same genus, as seen in Anaerotignum and Angelakisella.

Discussion

We conducted an in-depth investigation of the gut microbiome of A. gazella, primarily through shotgun metagenomic sequencing and complemented by 16S rRNA amplicon sequencing. This integrated approach provided valuable insights into microbial diversity, metabolic pathways, and ecological adaptations of the gut microbiome. The present work is particularly significant in view of the unique dietary habits, health status, and environmental challenges encountered by A. gazella.

The consistent detection of Arthropoda and Malacostraca DNA fragments in fecal metagenomics, coupled with physical observations of krill residues and diagnostic coloration, confirms the central role of Antarctic krill in the species’ diet [3, 76]. Given that chitin constitutes the primary residual component of krill after digestion by A. gazella, its further degradation by the gut microbiome is a direct reflection of the host’s diet. Indeed, chitinase and related chitin utilization pathways were detected across all metagenomic samples. Importantly, moving beyond community-level functional abundance, our MAG-resolved analysis directly links specific taxa to chitin degradation steps: Faecousia, Massiliimalia, and Angelakisella carry chitinase genes, positioning them as putative primary degraders responsible for the initial hydrolysis of polymeric chitin into soluble oligosaccharides, though direct evidence for chitin utilization by these genera remains undocumented. By contrast, Clostridium and Holdemania harbor the greatest diversity of chitin-degradation-related enzymes yet lack chitinase and lytic polysaccharide monooxygenase, suggesting that they likely act as secondary degraders, relying on other community members to cleave chitin into oligomers for subsequent metabolism. Nevertheless, previous studies have reported that certain Clostridium species encode chitinases [77, 78]; their absence in our dataset likely reflects limitations of shotgun metagenomic sequencing or assembly quality rather than a true biological deficiency. The three Helicobacter MAGs contribute only a small repertoire of chitin-related genes, all annotated as sugar transporters (TC.SSS-type permeases, peptide ABC transporters, or PTS systems), indicating a role restricted to oligosaccharide uptake rather than direct enzymatic hydrolysis. Taken together, these MAG-resolved data demonstrate that no single genome encodes the complete chitin degradation pathway. Complete depolymerization of chitin to GlcNAc and its subsequent entry into central metabolism therefore appears to depend on sequential cross-feeding among taxa with complementary enzyme sets, a model that remains to be validated experimentally.

Beyond the traditional hydrolytic strategy for chitin degradation, our data suggest the potential coexistence of an oxidative degradation mechanism for chitin, even though the key enzyme LPMO was not identified. This could be owing to the low sequence homology of the respective gene to currently characterized homologs, especially potentially novel LPMO families [79, 80]. Importantly, the integration of shotgun metagenomics represents the first comprehensive effort to map both hydrolytic and oxidative chitin degradation pathways in Antarctic marine mammals, thus providing a robust framework for future investigations [72, 81]. Thus, our genomic findings not only corroborate established ecological knowledge but also provide a mechanistic, gene-centric validation of this critical trophic link.

Beyond the dominant krill signature, the metagenomic data revealed traces of additional eukaryotic material, including the terrestrial legume genus Lupinus, which is not known to be part of the natural diet of Antarctic fur seals. Several non-mutually exclusive hypotheses may explain the presence of Lupinus. First, Antarctic fur seals can occasionally range into sub‑Antarctic islands or coastal areas of southern South America where Lupinus naturally occurs [82, 83], raising the theoretical possibility that seals or their prey could be exposed to terrestrial plant material in coastal ecosystems. However, because A. gazella is an obligate carnivore, any such exposure would require indirect trophic transfer (e.g., via prey that had consumed detritus or plankton associated with terrestrial organic matter) [84]. Second, taxonomic assignment based on alignment against the NCBI-NR database may be confounded by uneven phylogenetic coverage and incomplete reference sequences, resulting in annotation bias and potential misclassification of distantly related taxa [85]. Third, contamination during sampling and downstream processing cannot be excluded, althought strict precautions were implemented. For instance, Lupinus pollen may be dispersed to the Fields Peninsula from Antarctic-proximal regions via long-distance aerosol transport [86], potentially introducing exogenous material during field collection. We therefore recommend that future studies include full‑process negative controls. Likewise, other identified plant taxa, including species associated with human agriculture and remote wild flora, should be interpreted with similar caution.

Turning to prokaryotic taxa, the detection of polar-adapted genera such as Psychrobacter, Polaromonas, and Deinococcus should also be carefully interpreted. These are ubiquitous in Antarctic environments, including soil, seawater, and sea ice [57, 87, 88]. In exposed field sampling settings, these psychrophiles can readily contaminate fecal samples through substrate contact or aerosol exposure [89, 90], or alternatively adhere to prey surfaces and be ingested via the food chain [91]. Therefore, the presence of these taxa may reflect environmental microbial background rather than established gut colonization, which is a limitation inherent to field-based sampling in exposed polar habitats [89]. Collectively, these findings demonstrate the value of seal fecal metagenomics as a tool for ecological forensics, offering a non-invasive window into both direct and indirect trophic interactions, with implications extending from Antarctic to global marine food webs. However, it is important to consider the contamination inherent to field-based sampling, and potential limitations in data analysis, such as sample sequencing and bioinformatics analysis, which may lead to errors in species annotation.

The overall microbial community composition exhibited a consistent pattern across the four individuals. However, given the extremely limited sample size (n = 4), this consistency should not be interpreted as evidence of a stable core microbiome, which would require much larger cohorts to distinguish truly persistent taxa from stochastic presence. Instead, the observation should be viewed as preliminary evidence of taxonomic overlap among a small number of conspecific individuals sharing similar diet and habitat. Bacteria account for over 94.5% in all samples, which aligns with previous studies on carnivorous marine mammals that bacterial taxa typically dominate due to their roles in nutrient metabolism and host-microbe symbiosis, and protein-rich diet likely reduces the prevalence of archaeal populations [92, 93]. The predominance of Bacillota in the gut microbiota of Antarctic fur seals, as revealed by both 16S and metagenomic analyses, thereby supports its recognized status as a dominant phylum in carnivorous marine species [17, 21, 92]. Clearly, differences among the samples at genus level were more pronounced than at phylum level by two methods. For example, in contrast to 16S rRNA amplicon data which identified Clostridium as the dominant genus, metagenomic data revealed Fusobacterium to be the most abundant. Nevertheless, both genera contribute to protein metabolism, a key process for hosts consuming krill, which is rich in lipids and proteins.

The observed discrepancy between 16S rRNA amplicon sequencing and shotgun metagenomic profiling is attributable to the distinct technical biases inherent in each approach. First, 16S rRNA gene copy number variation systematically distorts abundance estimates in amplicon sequencing. Members of the phylum Bacillota (averaging > 5 copies per genome, with some species harboring up to 15 copies; e.g., Clostridium contains 7–11 copies) contribute disproportionately more PCR templates than single-copy taxa [94, 95], artificially inflating their apparent abundance. Second, primer amplification bias further compounds the distortion of 16S amplicon results. Although the widely used 338F/518R primer pair is considered broadly conserved, no single universal primer set can achieve complete taxonomic coverage. Sequence mismatches at primer binding sites can compromise amplification efficiency for certain lineages, thereby introducing potential biases into amplicon-based community profiles [96, 97]. In contrast, shotgun metagenomics circumvents PCR- and primer-associated biases by directly sequencing total community DNA and inferring abundance from average genome coverage or single-copy marker genes, providing a more accurate representation of taxonomic composition [98]. Nevertheless, shotgun approaches are not without limitations: taxonomic assignment relies on reference database completeness, and variations in genome size and assembly efficiency between taxa can introduce compositional bias [99]. In this study, phylum-level concordance between the two methods supports the reliability of the overall community structure, whereas genus-level divergence highlights how methodological choice directly shapes perceived community composition at finer taxonomic resolution. Collectively, these results underscore that integrating 16S rRNA and shotgun metagenomics enables a more comprehensive and reliable characterization of complex microbial communities.

A recent virome study of Antarctic fur seals [20], identified 11 viral families, with the results showing that the largest proportion of contigs was associated with bacteriophage (40.8%), the remainder were associated with mammalian (21%), invertebrate (11.1%), fish (7.5%), and bird-infecting (6.1%) viruses. Similarly, in our study the viral community was dominated by the phylum Uroviricota, primarily represented by the class Caudoviricetes (bacteriophages), which accounted for 90.2% ~ 95.2%, and only a small fraction was classified as Herpesviridae, a family known to infect pinnipeds. Furthermore, only two viral families were shared between the two datasets—Guelinviridae (including five genera) and Microviridae (represented only by an unclassified Microviridae genus). Differences between the two studies likely stem from multiple factors and warrant further investigation. Previous studies have found that phages are the main component of animal gut viruses [100, 101], where they play a crucial role in maintaining the ecological balance of the gut microbiota [102], thus contributing to a healthy microbial community structure and host-beneficial functions [103].

Nevertheless, pathogenic risks were evident across multiple domains of the gut ecosystem. Beyond the core virome, we identified several sequences of concern. Notably, among the 20 most abundant viral genera, Lentivirus was detected in sample S59. Lentiviruses (including HIV, SIV, and BIV) are characterized by high host specificity, the capacity to cause immune dysfunction, and a recognized risk of cross-species transmission [104], potentially indicating underlying health complications in this individual. Concurrently, pathogenic bacteria were detected in our samples, including the pervasive presence of Campylobacterota, a phylum associated with gastrointestinal pathogens [105]. This finding aligns with the earlier detection of Campylobacter insulaenigrae and Campylobacter lari in Antarctic fur seals at Deception Island over a decade ago [106], confirming the established presence of these indigenous pathogens in the Antarctic region. Pathogenic potential was also evident at the MAGs level: specifically, S62-bin.1 clustered near H. trogontum and H. bilis, both of which are pathogenic bacteria isolated from the intestines of rodents [74, 75]; meanwhile, S62-bin.7 and S63-bin.4 clustered with W. succinogenes, a symbiont in the bovine rumen holding an extensive repertoire of genes homologous to known bacterial virulence factors [107] and has previously been detected in the gut of an Australian Sea Lion (Neophoca cinerea) with gastritis [108]. Further elevating the risk profile, we detected high proportions of the eukaryotic parasites Toxocara and Trichuris. This indicates the potential intestinal colonization by zoonotic parasites, a phenomenon which is likely exacerbated by climate-driven shifts in intermediate host distributions. Moreover, the identification of virulence factors, including Type IV secretion systems (T4SS) and lipopolysaccharides (LPS), also highlights potential pathogenic risks within the gut microbiome [109, 110]. The increasing human activity in Antarctica for research and tourism heightens concerns regarding the potential for cross-species transmission of such pathogens [25, 111, 112].

Of equal concern is the prevalence of antibiotic resistance genes (ARGs) potentially conferring resistance to tetracycline, vancomycin, and multidrug. The exceptionally high abundance of multidrug resistance genes is alarming; it serves as the symbol of the entry into the “post-antibiotic era” and is currently one of the top ten global health threats. The high presence of vancomycin resistance genes is particularly worrisome, as vancomycin represents a last-resort antibiotic for Methicillin-Resistant Staphylococcus aureus (MRSA) infections [113, 114]. Compelling evidence indicates that antibiotic treatments significantly alter microbiota composition [115–117], and the overuse of antibiotics may accelerate the accumulation of resistance genes within microbiomes [118]. Up to 80% of isolates from wildlife (e.g., avian and marine mammal species) showed antibiotic resistance from Antarctic Peninsula [17]. Long-term exposure of Antarctic fur seals to these substances could elevate disease risks, threatening population viability. However, the local geochemical background must be carefully taken into account when interpreting these antimicrobial resistance gene data. The soils of King George Island can be influenced by regional volcanic activity (including historical events from volcanoes such as Deception Island), which may introduce heavy metals [119–122]. It is now well established that heavy metal contamination can drive the maintenance and spread of antibiotic resistance genes through co-selection mechanisms, even in the absence of antibiotic exposure [123]. Given that our sampling site on the Fildes Peninsula is situated within this geochemical context, and that Antarctic fur seals are known to bioconcentrate mercury and other heavy metals in their excreta [9, 124], we cannot exclude the possibility that a portion of the detected ARGs is maintained through co-selection mechanisms. Consequently, while our findings highlight the presence of clinically relevant ARGs in this remote host population and underscore the need for continued environmental surveillance, the relative contributions of geochemical metal exposure versus direct anthropogenic antibiotic selection in shaping the resistome remain to be disentangled. Importantly, genomic detection of ARGs indicates only the potential for resistance. Definitive confirmation of clinically relevant resistance requires phenotypic susceptibility testing. Thus, our data identify ARG reservoirs in this remote ecosystem, but do not demonstrate functional resistance or ongoing antibiotic selection pressure.

These observations nevertheless emphasize Antarctic ecosystems as sentinel sites for monitoring emerging One Health threats. Future research should integrate metagenomic surveillance with contaminant tracing to distinguish natural microbial adaptations from human-driven perturbations.

Conclusion

Integrating shotgun metagenomics with 16S rRNA amplicon sequencing, our results provide the first comprehensive investigation of the A. gazella gut microbiome. Our analysis confirms its krill-based dietary adaptions and the corresponding chitinolytic capacity of its gut microbiota, while also revealing a concurrent and concerning presence of enteric pathogens (bacteria, viruses, and eukaryotic parasites) alongside diverse antibiotic resistance genes that indicate potential antibiotic resistance or co-selection mechanisms. These findings underscore the metabolic adaptability of the gut microbiome to polar conditions, while also highlighting growing threats from anthropogenic influences such as antibiotic exposure, human activity, and climate change. The detection of virulence factors and zoonotic agents further suggests potential health risks to seal populations and raises broader One Health concerns in Antarctic ecosystems. These insights are pivotal for informing conservation strategies and monitoring ecosystem health in rapidly changing Antarctic environments to help restore ecological balance while safeguarding vulnerable polar species. The most critical limitation of this study is the small sample size (n = 4), which precludes population-level, ecological, or epidemiological generalizations. The shared taxa and functions identified across these four individuals represent observational overlap rather than a statistically validated core microbiome. Polar field research is conducted under extreme environmental constraints, making biological specimens inherently scarce, exceedingly difficult to collect, and thus severely limiting achievable sample sizes. Consequently, future studies with substantially larger sample sizes are needed to validate these preliminary findings and to robustly define the core microbiome of Antarctic fur seals. Overall, this work establishes a foundational microbiome profile for A. gazella and emphasizes the urgency of continuous ecosystem monitoring to support the conservation of Antarctic wildlife in an era of rapid environmental change.

Author contributions

Tingyi Lai performed most of the data analysis and drafted the paper with Yisi Liu and Zedong Duan. Li Liao supervised the study and revised the paper. Shiyuan Su collected the samples. Yufei Dai and Minhao Gao conducted part of experiments. Mukan Ji and Haitao Ding reviewed the manuscript.

Funding

National Key Research and Development Program of China, 2022YFC2807501, National Natural Science Foundation of China, 42476264

Data availability

All data produced or analyzed in this study are provided in the article and its supplementary information. The raw 16S rRNA and metagenomic data reported in this study have been uploaded to NCBI’s Sequence Read Archive under the BioProject accession number: PRJNA1390320 and PRJNA1392086.

Declarations

Competing interests

The authors declare no competing interests.